SeKondBrain
Sign Up
Our research

Research at the
frontier of intelligence.

We don’t just build on AI research — we contribute to it. Our work argues for a new architecture for how machines represent and reason with knowledge.

Read our latest paper
The research problem

Scale doesn’t
solve meaning.

The industry is racing to make models bigger and faster. We think the harder and more important problem is how knowledge is represented, recalled and used in the first place.

Most AI today predicts the next most probable token in a sequence — extraordinary at it, and so fluent the output feels like understanding. But fluency isn’t understanding. Models work brilliantly at the token level and retrieval works over whole documents, yet neither aligns with the level at which knowledge actually lives: the concept.

A single document holds many concepts; a single concept is scattered across many documents. That mismatch is architectural — not something more parameters or a bigger context window will patch. It’s a research problem, and one that’s been developing in academic work for over a decade. That’s the problem we’ve taken on.

Atomic Units of X: The Compression Layer of Intelligence

May 2026 · Duggal, Vasileiadis & Pradyumna S R

The formal core of the programme.

Intelligence — human or artificial — operates via atomic units that function as compression layers, dynamically composed into novel configurations. The paper introduces the Compression Calculus, formalising representational efficiency across ten domains, and the Compounding Cascade — compression ratios of 10× to over 10,000× per domain that multiply across abstraction layers. It reframes LLMs as dynamic fusion engines over atomic units, and points to self-generating systems that discover new atoms through compression-driven library learning.

What we found: concept-level representation cut message length by 46.2% while preserving meaning · reconstruction fidelity passed 100% of the time, zero critical failures · atomic retrieval hit 100% Recall@5 against 91% for chunk retrieval, using 47.5% less retrieved context · and reliable composition of atoms remains unsolved (F1 0.13) — named openly as the next problem.

Special thanks to Benjamin Brey, Dr Sharon Jheeta and Priyanka Kocchar.

Read the full paper

Cognition as an Ecology: Designing AI for Internal Memory, Proximal Ecosystems, and External Unknowns

December 2025 · Sachin Dev Duggal

Human knowledge lives in three domains: Internal — what you hold yourself; Proximal — the accessible resources around you, colleagues and familiar systems; and Frontier — the external unknowns. The paper argues that creativity and innovation arise from optimal synthesis between the three — the surprise gradient — and that an AI architecture which collapses them into one retrieval problem fails at all of them. This is the frame everything we build sits inside.

Read the full paper
The wider frontier

We’re not
working alone.

Our approach sits within a growing body of research — neurosymbolic AI, cognitive science and knowledge representation all point toward structure and meaning, not scale alone.

The idea that intelligence depends on reusable, composable units isn’t ours alone — it runs through decades of research. Cognitive science shows that experts think in compressed chunks, not raw detail. Information theory ties compression to prediction and understanding.

Neurosymbolic AI works to combine the fluency of neural models with the structure and traceability of symbolic reasoning. And recent work on compositional generalisation and library learning shows systems discovering and reusing their own abstractions.

Our contribution is to bring these threads together into one measurable framework — and to build on it. We’re advancing a frontier alongside a scientific community increasingly converging on the same insight.

Where we stand

We’ll tell you exactly where we are.
And where we’re going.

Type 2.5 today. The next step is the one worth arguing about.

In 2020, Henry Kautz set out a taxonomy for how neural and symbolic systems can be combined — a ladder running from systems that merely bolt the two together, through those where each calls the other, to systems where symbolic reasoning is genuinely embedded in the neural substrate.

Most of what ships today sits near the bottom of that ladder: a language model with a retrieval step attached. SeKondBrain sits at Type 2.5 — the concept graph is not a lookup the model calls, it is structure the reasoning runs through. Meaning is held symbolically and stably; reasoning is traceable; the neural layer reads the world and the symbolic layer holds what it means.

We are not at the top of the ladder, and we will not claim to be. The step above is where symbolic structure and neural computation stop being two systems in conversation and become one. That is the problem we are working on — and the next section is where we are with it.

Read the paper
What’s next

The problems we
haven’t solved yet.

Three active workstreams. One has data; none has a victory lap.

In draft
Creativity on SeKondBrain
If knowledge decomposes into atoms, creativity is what happens when distant ones fuse. This white paper — in draft — argues that novelty can be made systematic: autonomous agents fusing semantically distant concepts over a concept graph, then evaluating what the fusion produces. It closes the loop the ecology paper opened: the surprise gradient, made operational.
Workstream
Concept-Grounded Attention (CGA)
Our published work established that meaning can be held at the level of concepts, and that retrieval over them is sharper and lighter. What it did not solve is composition — reliably assembling atomic concepts into more complex ones. We reported that openly (F1 0.13) and named it as the next problem. CGA is our line of attack: grounding attention itself in the concept graph, so a model attends over stable meaning rather than surface tokens. It’s early, and we’re claiming nothing yet.
Workstream
Context compaction
If intelligence is compression, memory should get cheaper and sharper as it matures — not heavier. Our compaction research studies exactly that, on operational data from Kemory: a tiered context policy that distils raw memory through layered summaries into a compact model of the user, deduplicating as it goes. The study turns the architecture’s compression expectation into a measured result. Paper in preparation.
Follow the work
Our contribution

Research in the open,
for the field.

We share our thinking with the research community because a new architecture for intelligence is bigger than any one company.

We take the science seriously — real research and real-world application, not a demo that falls apart on contact with reality. And we believe the shift we’re describing is too important to keep behind a curtain. So we publish: our frameworks, our findings, and the problems we haven’t solved yet. Some of it is settled; some is still open. That’s how a frontier moves. If you’re a researcher, a builder or an investor who cares about where intelligence goes next, this is the work — and the conversation — we’d like you to be part of.

The research

Read the work.
Follow the frontier.

Read the work, or follow along as the next results land. This is the science behind everything we build — read the papers in full, see how the thesis becomes a product on the Platform, or follow the research as it grows with the Community.

Read the papersJoin the community